baseline_v1/src/make_figures.py

from pathlib import Path
import json,math
import numpy as np
import matplotlib
matplotlib.use('Agg')
import matplotlib.pyplot as plt
from matplotlib.colors import ListedColormap,BoundaryNorm
from scipy.stats import beta
R=Path(__file__).resolve().parents[1];F=R/'figures';F.mkdir(exist_ok=True)
plt.rcParams.update({'font.size':11,'axes.titlesize':12,'axes.labelsize':11,
 'figure.dpi':160,'savefig.dpi':210,'axes.spines.top':False,'axes.spines.right':False,
 'font.family':'DejaVu Sans'})
blue='#186B8C';orange='#CF763C';grey='#8B929C';dark='#243344'
a=np.genfromtxt(R/'results/access_redundancy.csv',delimiter=',',names=True)
fig,axs=plt.subplots(1,2,figsize=(9,3.5))
axs[0].plot(a['b'],a['exponent'],color=orange,lw=2.4,label='Supply distance grows')
axs[0].plot(a['b'],a['channel_exponent'],color=blue,lw=2.4,label='Local supply maintained')
axs[0].set(xlabel='Redundant subcomponents b',ylabel='Reliability exponent E',title='Redundancy can starve repair',xlim=(0,1100),ylim=(0,190))
axs[0].axhline(math.log(1e6/.05),color=grey,ls='--',lw=1)
axs[0].text(570,19,'Example required exponent',fontsize=8,color=grey)
axs[0].legend(fontsize=10,loc='upper left')
axs[1].plot(a['b'],a['raw_error'],color=orange,lw=2.4,label='Conversion-inclusive error')
axs[1].axhline(.2,color=dark,ls='--',label='Assumed tolerance = 0.20')
axs[1].set(xlabel='Redundant subcomponents b',ylabel='Raw subcomponent error',title='Depletion closes the window',xlim=(0,1100),ylim=(0,.8))
axs[1].legend(fontsize=10)
fig.text(.5,.01,'Illustrative analytical model; no chemical rates or device tolerance were measured.',ha='center',fontsize=8,color=grey)
fig.tight_layout(rect=(0,.055,1,1));fig.savefig(F/'reliability_access.png');plt.close(fig)

cmap=ListedColormap(['#CA4560','#F0F1F2','#7387A2','#E2A844'])
norm=BoundaryNorm([-1.5,-.5,.5,1.5,2.5],4)
fig,axs=plt.subplots(2,3,figsize=(8.2,5.7))
for row,dim in enumerate([2,3]):
    ref=np.load(R/f'results/snapshot_{dim}d_reference.npz')
    no=np.load(R/f'results/snapshot_{dim}d_no_repair.npz')
    vals=[ref['target'],ref['final'],no['final']]
    for col,v in enumerate(vals):
        if dim==3:v=v[:,:,v.shape[2]//2]
        axs[row,col].imshow(v.T,origin='lower',cmap=cmap,norm=norm,interpolation='nearest')
        axs[row,col].set_xticks([]);axs[row,col].set_yticks([])
        if row==0:axs[row,col].set_title(['Target','Repair + delayed lock','No repair'][col])
        if col==0:axs[row,col].set_ylabel('2-D device' if dim==2 else '3-D central slice')
fig.text(.5,.025,'Light: void / sacrificial role    Blue: support role    Gold: recruited phase',ha='center',fontsize=9)
fig.tight_layout(rect=(0,.055,1,1));fig.savefig(F/'growth_snapshots.png');plt.close(fig)

summary=json.loads((R/'results/growth_summary.json').read_text());labels=['reference','no_repair','early_lock','no_internal_supply','common_mode','conversion_damage']
short=['Reference','No repair','Early lock','Boundary\nsupply','Wrong\nreference','Conversion\ndamage']
fig,axs=plt.subplots(1,2,figsize=(8.2,4.1),gridspec_kw={'width_ratios':[1.3,1]},sharey=True)
x=np.arange(6)
short=['Reference','No repair','Early lock','Boundary supply','Wrong reference','Conversion damage']
for dim,color,offset in [(2,blue,-.18),(3,orange,.18)]:
    vals=[next(r for r in summary if r['dim']==dim and r['case']==lab) for lab in labels]
    axs[0].barh(x+offset,[v['material_fidelity_mean'] for v in vals],height=.34,color=color,label=f'{dim}-D',
      xerr=[v['material_fidelity_sd'] for v in vals],capsize=2,error_kw={'lw':1})
axs[0].set_yticks(x,short,fontsize=10);axs[0].set_xlim(.7,1.005);axs[0].set_xlabel('Material fidelity');axs[0].set_title('Local repair improves labels');axs[0].legend(fontsize=10,loc='lower left')
vals=[next(r for r in summary if r['dim']==2 and r['case']==lab) for lab in labels]
axs[1].barh(x,[v['functional_passes'] for v in vals],color=[blue]+[grey]*5,height=.6)
axs[1].set_xlim(0,8.6);axs[1].set_xlabel('Functional passes / 8');axs[1].set_title('Function remains difficult')
axs[0].invert_yaxis()
fig.text(.5,.012,'Coarse stochastic model. Error bars: between-run SD; eight runs per condition.',ha='center',fontsize=9,color=grey)
fig.tight_layout(rect=(0,.075,1,1));fig.savefig(F/'ablation_results.png');plt.close(fig)

stress=json.loads((R/'results/transport_stress.json').read_text());fig,axs=plt.subplots(1,2,figsize=(8,3.5))
for i,spacing in enumerate([5,0]):
    vals=[v for v in stress if v['config']['channel_spacing']==spacing]
    for ax,key in zip(axs,['completed_fraction','material_fidelity']):
        yy=np.array([v[key] for v in vals]);ax.bar(i,yy.mean(),color=[blue,orange][i],alpha=.8)
        ax.scatter(i+np.linspace(-.1,.1,len(yy)),yy,color=dark,s=19,zorder=3)
for ax,title in zip(axs,['Completed fraction','Material fidelity']):
    ax.set_xticks([0,1],['Internal feed planes','Boundary supply only'],fontsize=9)
    ax.set_ylim(0,1.);ax.set_title(title);ax.axhline(1,color=grey,ls='--',lw=1)
fig.text(.5,.015,'Four runs each; both conditions fail to complete by the declared deadline.',ha='center',fontsize=8,color=grey)
fig.tight_layout(rect=(0,.075,1,1));fig.savefig(F/'transport_stress.png');plt.close(fig)

nums=json.loads((R/'results/numerical_summary.json').read_text());mc=nums['monte_carlo']
fig,ax=plt.subplots(figsize=(6.5,3.8));b=np.array([v['b'] for v in mc]);yy=np.array([v['empirical'] for v in mc])
lo=np.array([beta.ppf(.025,v['failures'],v['trials']-v['failures']+1) if v['failures'] else 0 for v in mc])
hi=np.array([beta.ppf(.975,v['failures']+1,v['trials']-v['failures']) if v['failures']<v['trials'] else 1 for v in mc])
ax.errorbar(b,yy,yerr=[yy-lo,hi-yy],fmt='o',color=orange,capsize=3,label='Monte Carlo, exact 95% interval')
ax.plot(b,[v['exact'] for v in mc],color=blue,label='Exact binomial tail')
ax.plot(b,[v['chernoff'] for v in mc],color=grey,ls='--',label='Chernoff upper bound')
ax.set_yscale('log');ax.set(xlabel='Redundancy b',ylabel='Module-failure probability',title='Numerical check of the restricted probability model')
ax.legend(fontsize=10);fig.tight_layout();fig.savefig(F/'binomial_check.png');plt.close(fig)
print('Saved five reproducible figures')